CHAPTER 21 Location Updates for Efficient Routing in Ad Hoc Networks IVAN STOJMENOVIC ´ DISCA, IIMAS, UNAM, Universidad Nacional Autonoma de Mexico 21.1 INTRODUCTION Mobile ad hoc networks consist of wireless hosts that communicate with each other in the absence of a fixed infrastructure. Some examples of the possible uses of ad hoc network- ing include soldiers on the battlefield, emergency disaster relief personnel, and networks of laptops. Sensor networks are a similar kind of network that have recently been investi- gated. Nodes in a sensor network are lighter, computationally less powerful, and more likely to be static compared to nodes in an ad hoc network. Hundreds or thousands of such nodes may be placed to monitor and control a physical environment from possibly remote locations. These nodes frequently switch their activity status to preserve battery power, which poses additional challenges for the design of efficient data collection algorithms. Ad hoc and sensor networks are self-organized and collaborative. Zero configuration net- working is also required for environments in which administration is impractical or im- possible, such as home or small offices or embedded systems “plugged together” (as in an automobile), or for allowing impromptu networks between the devices of strangers on a train [50]. In this chapter, we consider the routing task in which a message is to be sent from a source node to a destination node. Due to propagation path loss, the transmission radii are limited. Thus, routes between two hosts in a network may consist of hops through other hosts in the network. The task of finding and maintaining routes in the network is nontriv- ial since host mobility causes frequent unpredictable topological changes. “Sleep period operation” (when some nodes become temporarily inactive) poses additional challenges for routing protocols. Macker and Corson [35] listed qualitative and quantitative independent metrics for judging the performance of routing protocols. Desirable qualitative properties include: distributed operation, loop-freedom (to avoid the worst-case scenario of a small fraction of packets spinning around in the network), demand-based operation, and “sleep period operation.” Some quantitative metrics that are appropriate for assessing the performance of any routing protocol include [35] end-to-end data delay and average number of data bits (or control bits) transmitted per data bits delivered. Our review (with primary interest in 451 Handbook of Wireless Networks and Mobile Computing, Edited by Ivan Stojmenovic´ Copyright © 2002 John Wiley & Sons, Inc. ISBNs: 0-471-41902-8 (Paper); 0-471-22456-1 (Electronic) location-based techniques) indicates that most proposed routing algorithms (more precise- ly, their performance evaluations) ignore one or more of these important metrics. Ad hoc networks are best modeled by minpower graphs constructed in the following way. Each node A has its transmission range t(A). Two nodes A and B in the network are neighbors (and thus joined by an edge) if the Euclidean distance between their coordinates in the network is less than the minimum between their transmission radii (i.e., d(A, B) < min {t(A), t(B)}. If all transmission ranges are equal, the corresponding graph is known as the unit graph. In the unit graph model, forwarded messages simultaneously provide ac- knowledgments for received messages. The minpower and unit graphs are valid models when there are no obstacles in the signal path (e.g., a building). Ad hoc networks with ob- stacles can be modeled by subgraphs of minpower or unit graphs. Most papers use unit graphs for the performance evaluation of proposed routing protocols. In the next section, we classify existing routing algorithms according to a number of criteria. This section will also review a number of routing protocols. In Section 21.3, loca- tion updates between neighboring nodes are discussed. Sections 21.4 through 21.7 de- scribe several existing location update methods. Priority is given to newer algorithms with novel approaches. Performance evaluation issues are discussed in Section 21.8. The refer- ence section gives an extensive list of relevant articles. 21.2 CLASSIFICATION OF ROUTING ALGORITHMS We shall now review the main characteristics of proposed routing algorithms in light of desired qualitative and quantitative properties [35] and a few additional characteristics. 21.2.1 Demand-Based Operation Routing algorithms can be classified as proactive or reactive. Proactive protocols maintain routing tables when nodes move, independently of traffic demand, and thus may have un- acceptable overhead when data traffic is considerably lower than mobility rate. For in- stance, shortest- (weighted) path-based-solutions [3, 43, 45] are too sensitive to small changes in local topology and activity status (the later even does not involve node move- ment). The communication overhead involved in maintaining global information about the networks is not acceptable for networks whose bandwidth and battery power are severely limited. They are not elaborated on further in this chapter. Routes in reactive algorithms are established when they are needed, in order to mini- mize the communication overhead. They are adaptive to “sleep period” operation, since inactive nodes simply do not participate at the time the route is established. One of well- known reactive algorithms is the source-initiated, on-demand routing strategy [5, 22, 41, 44, 45]. In this strategy, the source or intermediate node S issues destination search re- quest if the route to destination D is not available. The destination search is performed by flooding a “short” search message, so that each node in the network is reached. Flooding algorithms that reduce the number of retransmissions are surveyed and discussed in [50]. The path to destination is memorized in the process [5, 22, 41, 44, 45]. A variant of this strategy is proposed in [53]. Several search “tickets” (each ticket is a “short” message con- 452 LOCATION UPDATES FOR EFFICIENT ROUTING IN AD HOC NETWORKS taining the sender’s ID and location, the destination’s ID and best-known location and time when that location was reported, and a constant amount of additional information) that will look for the exact position of the destination node are issued by source S. When the first ticket arrives at the destination node D, D will report back to the source with a brief message containing its exact location, and possibly create a route for the source (the sec- ond phase). In the third phase, the source node then sends a full data message (“long” message) toward the exact location of destination. The efficiency of destination search de- pends on the corresponding location update scheme. A quorum-based, home-agent-based, and depth-first-search-based destination search and corresponding location update schemes are being developed [49, 53, 54]. Other location update and destination search schemes may be used, including an occasional flooding. If the routing problem is divided as described, the mobility issue can be algorithmically separated from the routing issue, allowing the application of routing algorithms with known destination in the second and third phases of the protocol. The choice is justified whenever the destination does not move significantly between its detection and message delivery, and information about neighboring nodes is regularly maintained. In the described approach, the communication overhead of routing algorithm is divided into the following components: location updates, destination searches (performed in accordance to location update scheme), and path cre- ation (or reporting from destination back to source). An interesting compromise between proactive and reactive methods is proposed in [7]. The algorithm is destination-initiated: a destination initiates a global path computation to itself using dynamic link metrics, which include a measure of “hotness” of the particular destination and congestion in the vicinity of the destination. The updated routes are short- est cost path routes, where queue length at each link (which is proportional to delay) is taken as the cost measure. 21.2.2 Distributed Operation We shall divide all distributed routing algorithms into localized and nonlocalized. Local- ized algorithms [12] are distributed algorithms that resemble greedy algorithms in which simple local behavior achieves a desired global objective. In a localized routing algorithm [4, 6, 14, 28, 29, 47–49, 55], each node makes the decision of which neighbor to forward the message based solely on the location of itself, its neighboring nodes, and the destina- tion. Although neighboring nodes may update each other’s location whenever an edge is broken or created, the accuracy of destination location is a serious problem. In some cases such as monitoring the environment by sensor networks, the destination is a fixed node known to all nodes (i.e., monitoring center). Localized algorithms are directly applicable in such environments. Otherwise, they may use destination search as the first step, routing short messages from destination to source as the second, and, finally, routing full message from source to destination. Localized routing algorithms that guarantee delivery [6, 11] (assuming that the destination location is accurate and message transmissions by nodes on the route do not collide with other traffic) show that localized algorithms can nearly match the performance of shortest path algorithms. All nonlocalized routing algorithms pro- posed in the literature are variations of shortest weighted path algorithm [3, 5, 9, 22, 32, 41, 43]. Zone-based approaches combining shortest paths within a zone and interzonal 21.2 CLASSIFICATION OF ROUTING ALGORITHMS 453 destination searches or routing tables are elaborated in [23, 33]. In zone-based routing al- gorithms [23], nodes are divided into nonoverlapping zones. Each node only knows node connectivity within its own zone, and routing within the zone is performed directly. If the destination is outside the zone, one location request is sent to every zone to find the desti- nation. This seems to add significant overhead, indicating that combined requests in this planar interzonal graph should be designed instead. Additional problems may arise when nodes within the same zone are disconnected and neighboring zones are not reachable from all nodes within a zone. Thus, this promising protocol needs further development. A zone-based protocol that does not use location information of nodes is described in [21]. GRID protocol [33] selects one node in each grid or zone, and these nodes serve as the backbone for routing tasks. 21.2.3 Location Information Most proposed routing algorithms do not use the location of nodes, that is, their coordi- nates in two- or three-dimensional space, in routing decisions [5, 23, 44, 45]. The distance between neighboring nodes can be estimated on the basis of incoming signal strengths (if some control messages are sent using fixed power). Relative coordinates of neighboring nodes can be obtained by exchanging such information between neighbors. Alternatively, the location of nodes may be available directly by communicating with a satellite through GPS (Global Positioning System) if nodes are equipped with a small low-power GPS re- ceiver. We believe that the advantages of using location information outweigh the cost of additional hardware, if any. The distance information, for instance, allows nodes to adjust their transmission powers and reduce transmission power accordingly. This enables using power, cost, and power cost metrics [10, 43, 48] and corresponding routing algorithms [48] in order to minimize energy required per routing task and to maximize the number of routing tasks that a network can perform. Routing tables that are updated by mobile soft- ware agents modeled on ants are used in [8]. Ants collect and disseminate location infor- mation about nodes. 21.2.4 Single-Path versus Multipath Strategies There exist several multipath, full-message strategies in which each node on the path sends a full message to several neighbors that are best choices for all possible destina- tion positions [4]. There is significant communication overhead, and lack of guaranteed delivery can make this approach inferior to even a simple flooding algorithm. Clever flooding algorithms may use about half of the nodes only for retransmissions [50], which often matches the number of nodes participating in routing in this method. In ad- dition, flooding guarantees delivery and requires no prior location updates for improved efficiency. In [20], it was argued that flooding is the best routing method for very high mobility rates. Multipath methods [4, 29, 52] may be regarded as flooding that is re- stricted to the request zone and, as such, can be used for geocasting (in which a message is to be delivered to all nodes located within a region). A multipath algorithm that con- sists of several single paths is proposed in [47]. A single nonoptimal path, full-message strategy is proposed in [1]. A short message, multipath destination search, full-message, 454 LOCATION UPDATES FOR EFFICIENT ROUTING IN AD HOC NETWORKS optimal singlepath method was discussed above. The localized algorithms in this cate- gory will be briefly described below. Several GPS-based methods were proposed in 1984–1986 using the notion of progress. Define progress as the distance between the transmitting node and receiving node project- ed onto a line drawn from the transmitter toward the final destination. A neighbor is in the forward direction if the progress is positive; otherwise it is said to be in the backward di- rection. In the random progress method [38], packets destined toward D are routed with equal probability toward one neighboring node that has positive progress. In the NFP method [17], a packet is sent to the nearest neighboring node with forward progress. Taka- gi and Kleinrock [35] proposed the MFR (most forward within radius) routing algorithm, in which a packet is sent to the neighbor with the greatest progress. The method is modi- fied in [17] by proposing to adjust the transmission power to the distance between the two nodes. Finn [14] proposed a Cartesian routing method that allows choosing any successor node that makes progress toward the packet’s destination. The best choice depends on the complete topological knowledge. Finn [14] adopted the greedy principle in his simulation: choose the successor node that is closest to the destination. When no node is closer to the destination than the current node, the algorithm performs a sophisticated procedure that does not guaranty delivery. Recently, three articles [4, 28, 29] independently reported vari- ations of routing protocols based on direction of destination. In the compass routing (or DIR) method proposed by Kranakis, Singh, and Urrutia [28], the source or intermediate node A uses the location information for the destination D to calculate its direction. The location of one-hop neighbors of A is used to determine for which of them, say C, the di- rection AC is closest to the direction of AD. The message m is forwarded to C. The process repeats until the destination is, hopefully, reached. The GEDIR routing algorithm [47] is a variant of greedy routing algorithm [14] with a “delayed” failure criterion. GEDIR, MFR, and compass routing algorithms fail to deliver messages if the best choice for a node cur- rently holding a message is to return it to the previous node [47]. A GFG routing algo- rithm that guarantees delivery by finding a simple path between source and destination is described in [6]. It is based on constructing a planar subgraph (e.g., Gabriel graph) and providing routing in the planar subgraph that guarantees delivery. This procedure is called on whenever the greedy algorithm fails, and is recalled whenever a closer node (than the previously failing node) is encountered. The GFG algorithm [6] was implemented in [26] by including MAC layer considerations and location updates for experiments with moving nodes. The performance of the GFG algorithm was improved in [11] by adding a shortcut procedure and applying the internal node concept of Wu and Lee [57]. The hop count is very close to the hop count of the shortest path algorithm for dense graphs (below 20% excess hop count for graphs with average degrees Ն6) and about twice as long for sparse graphs. Corresponding power- and cost-aware routing algorithms with guaranteed deliv- ery are developed in [46]. 21.2.5 Loop Freedom Interestingly, loop freedom, a basic criterion of Macker and Corson [35] was neglected in many papers. GEDIR and MFR algorithms are inherently loop-free [47]. The proofs of this are based on the observation that distances (or dot products) of nodes toward the des- 21.2 CLASSIFICATION OF ROUTING ALGORITHMS 455 tination are decreasing. A counterexample showing that undetected loops can be created in directional-based methods [4, 28, 29] is given in [47]. The method is therefore not loop- free. The algorithms in [6, 11, 14, 35] and shortest-weighted-path-based routing schemes are loop-free. 21.2.6 Memorization of Past Traffic Most reported algorithms require some or all nodes to memorize past traffic as part of current the routing protocol or to memorize the previous best paths for providing future path to the same destination. Solutions that require nodes to memorize routes or partic- ular information about past traffic are sensitive to node queue size and changes in node activity and node mobility while routing is ongoing. One form of such memorization is provided by routing tables, which memorize the last successful path to each destination. Reduction in the size of routing tables (and, consequently, in the communication over- head to maintain them) was proposed in [25, 57] by defining backbone structures. Each node in the network is either in the virtual backbone or at most r hops away from a vir- tual backbone node. Clustering has frequently been used to provide such a backbone [25], where the r-cluster is defined as the set of all the nodes within distance at most r hops from a given node, referred to as the clusterhead of the r-cluster. Border nodes are nodes that belong to two or more clusters. Clusterheads are backbone nodes, and two “neighboring” backbone nodes may be up to 2r + 1 hops away. Thus, communication be- tween two backbone nodes may go through both backbone and nonbackbone nodes. A distributed scheme for initiating is based on selecting, repeatedly, a node with a maximal number of unassigned r hop neighbors as the backbone node, and assigning all its r hop neighbors to that node. Such a backbone is also used in the routing algorithm [30]. The maintenance of cluster structure is known to require significant communication overhead (for instance, local changes may cause global updates by the chain effect) [57]. A sig- nificantly better backbone structure, one that does not require any communication over- head and provides connectivity between nodes, is described in [57] and is based on sev- eral definitions of dominating sets. Localized routing algorithms discussed above [6, 14, 28, 46–48, 55] do not memorize past traffic at any node. The algorithms [4, 29, 52] require nodes to memorize past traffic to avoid infinite mutual flooding between neighboring nodes. Memorization of escape loops is needed in directional-based methods (alternatively, messages need to carry time- out stamps). In flooding GEDIR and MFR algorithms [47], messages are flooded at nodes in which basic algorithms fail, and these nodes refuse further copies of the same message. These algorithms guaranty delivery. Routing algorithms that use depth-first search (DFS) in the search for destination are discussed in [24, 49]. Memorization here is imposed by the DFS process. The algorithm guarantees delivery but the efficiency depends on the ac- curacy of the destination information. Memorization is needed in sensor networks for data fusion [12] to avoid multiple reports of the same information. Quality of service routing, in which the path needs to satisfy delay, bandwidth, and connection time criteria [49], re- quires that nodes memorize the QoS path; thus, using DFS for its construction does not impose any memorization overhead. 456 LOCATION UPDATES FOR EFFICIENT ROUTING IN AD HOC NETWORKS 21.3 LOCATION UPDATES BETWEEN NEIGHBORING NODES One of most important ingredients in all location update schemes is the update between neighboring nodes. The question is when does a node decide to send a message to all its neighbors announcing its new location. We shall review the methods used in literature. As a basic (or “bonus”) update, nodes may update their location information with each ex- change of routing messages between them. Karumanchi, Muralidharan, and Prakash [27] discussed the question when to update location, and argued that distance-based updates (based on absolute distance traveled since the last update) and movement-based updates (based on the velocities of nodes) may have limited usefulness in ad hoc networks (such location updates are used in [4, 29]). For instance, nodes may move within a small circle, causing unnecessary location updates. They concluded experimentally that the best strategy is to update when a certain prespeci- fied number of links incident on a node have been established or broken since the last up- date [27]. The basic update procedure is performed by each moving node whenever it observes that, due to its movement, an existing edge will be broken (that is, the distance between two nodes becomes >R). In order to minimize the number of location update messages, the message could be sent by only the node endpoint (of the broken edge) with greater speed. Similarly, the same action may be taken when a new neighbor is detected. New neighbor X may be detected after X transmitted its location update following an edge breakup with another node. Thus, new neighbors that receive such messages may then re- act by informing X about their presence. Alternatively, the creation of a new link can be detected if two-hop information is available to nodes. To decide whether an edge is made or broken, a node may use last available informa- tion about its direct neighbors and other nodes in the network. However, when all nodes are moving in the same direction (as in military or rescue missions), such a procedure may result in unnecessary updates. To reduce overhead in such scenarios, connection time is introduced as follows. The availability of GPS enables nodes to estimate the connection time with other nodes, as proposed in [49, 51]. The connection time is defined as the esti- mated duration of a connection between two neighboring nodes. Neighboring nodes fre- quently update their location to each other, and this information may be used to estimate the direction and speed of their movements. In turn, this suffices to estimate the connec- tion time. Let A and B be the two neighboring nodes that move at speeds a and b, respec- tively. Here, A and B are position vectors and a and b are directional vectors. At time t, they move to new positions AЈ = A + at and BЈ = B + bt. They will loose their connection when the distance between them becomes >R, where R is the radius of the corresponding unit graph (or the smaller of their transmission radii in case of minpower graphs). The time t at which the connection will be lost can be estimated by solving the quadratic equa- tion |AЈBЈ| = |B – A + (b – a)t| = R [49, 51]. When the time expires, the edge is assumed to broken and a location update is sent to all neighbors. Similar criteria can be used to esti- mate the time a connection will be made, and act accordingly. The variants of this basic update may include adjusting transmission radius to tR for some value of t, to reach more or fewer neighbors. Location updates are short messages, 21.3 LOCATION UPDATES BETWEEN NEIGHBORING NODES 457 and nodes may spend more energy for short messages, as suggested by Lin and Liu [32]. They discussed this difference and even proposed an extreme difference in radii for short and long messages. They found that nodes are able to send their new location to all other nodes in a network with a single broadcast (single-hop network for location updates). However, when sending exact data, the network is treated as a multihop one. Note that the single-hop location update broadcast may fail to reach a number of nodes due to obstacles in the field or presence of other transmissions. Another possible solution is to keep the transmission radius at R, but retransmit from each of the neighbors that are a few hops away. However, these retransmissions also require power (from neighboring nodes), and may cause the broadcast storm problem [NTCS]; therefore, their efficiency is doubtful. Note that a node that receives a location update aimed at known neighboring nodes, and discovers that it is now a neighbor of transmitting node that is not aware of it, will treat this event as the creation of a new edge and react by sending its own location update in response. This basic location update procedure may be used as a counter for “deeper” location updates. For instance, Basagni et al. [4] used parameter p (the distance traveled from the last update) as such a counter. Two-hop, four-hop, and flooding messages are sent on every first, second, and third counter, respectively. 21.4 REQUEST ZONE ROUTING A distance routing effect algorithm for mobility (DREAM) is described in [4]. The source or any intermediate node A calculates the direction of destination D and, based on the mo- bility information about D, chooses an angular range. The message m is forwarded to all neighbors whose direction belongs to the selected range. The range is determined by the tangents from A to the circle centered at D, with radius equal to a maximal possible move- ment of D since the last location update. The area containing the circle and two tangents is referred as the request zone [29]. Ko and Vaidya [29] described, independently, an almost identical algorithm, called the LAR Scheme 1, and a few modifications of it. The modifi- cations include sending route requests before the message itself [22]. Note that a route re- quest may be considered as a routing of short messages, as discussed above. Recovery procedures, based on partial or full flooding, to start flooding if the given algorithm fails to find the route within a timeout interval, are proposed in both papers [4, 29]. Ko and Vaidya [29] also proposed the LAR Scheme 2. In this scheme, the source or each interme- diate node A will forward the message to all nodes that are closer to the destination than A is. Wu and Harms [56] proposed to improve the location update part of the LAR algo- rithm. In [56], any two neighboring nodes periodically exchange full routing table (infor- mation about all nodes in the network). The definition of the request zone [4, 29] was modified in [52] in order to provide a uniform framework with the corresponding notions in GEDIR and MFR methods. Stoj- menovic [52] discusses the V-GEDIR, CH-MFR, and R-DIR methods, in which m is for- warded to exactly those neighbors that may be the best choices for a possible position of destination (using the appropriate criterion). The request zone in the R-DIR method [52] may include one or two neighbors that are outside of angular range, because they can have the closest direction for the tangents to the circle. In the V-GEDIR method, these neigh- 458 LOCATION UPDATES FOR EFFICIENT ROUTING IN AD HOC NETWORKS bors are determined by intersecting the Voronoi diagram of neighbors with the circle (or rectangle) of possible positions of destination; the portion of the convex hull of neighbor- ing nodes is analogously used in the CH-MFR method. The Voronoi diagram of n distinct points in the plane is a partition of the plane into n Voronoi regions, each one associated with each point. The Voronoi region associated with node A consists of all the points in the plane that are closer to A than to any other node. It can be shown that each region is a convex polygon (possibly unbounded) determined by bisectors of A and other nodes (more precisely, each region is the intersection of all such bisectors). It is well known that the Voronoi diagram for n points in the plane can be con- structed in O(n log n) time [40] and consists of O(n) line segments. Node S, currently holding a message for destination D, computes the Voronoi diagram of all its n neighbors. For example, in Figure 21.1, the Voronoi diagram of five neighbors A, B, C, E, and F is shown in dashed lines. Consider the circle (or other region) where des- tination D can be located. Different locations of D inside the circle correspond to different choices of closest nodes among A, B, C, E, and F. For each position of D, the closest node is the one whose Voronoi region contains the position. Thus, the nodes that are closest to some positions of destination are exactly those nodes whose Voronoi regions intersect the circle of possible destination positions (e.g., B and E in Figure 21.1). 21.5 DOUBLING CIRCLES ROUTING Amouris, Papavassiliou, and Lu [1] presented a position-based, multizone routing proto- col for wide area mobile ad hoc networks. Their algorithm is based on position updates within circles of increasing radii. Each node updates its location to all nodes located with- in a circle of radii P, 2P, 4P, 8P, . . . (each subsequent circle has a twice larger radius than 21.5 DOUBLING CIRCLES ROUTING 459 Figure 21.1 Voronoi diagram and the request zone. the previous one). Whenever a given node A moves outside one of these circles of radius 2 t P for some t, node A broadcasts its location update to all nodes located inside a circle centered at the current node position and with radius 2 t+1 P. The routing toward destination then follows these circles of last updates. Source nodes send messages toward the last re- ported position of destination (using the DIR method) that has moved within the circle of some radius since the last report. As routing the message moves closer to the destination, the information about position of destination becomes more precise, and nodes are able to send messages toward the centers of circles with twice smaller radii than previously, until the node is eventually reached. This process is illustrated in Figure 21.2. The source S sends a message toward DЈ, the last known position of destination D. The routing is later redirected toward newer position DЈ and finally to exact position D. This method is very interesting and certainly competitive. We observe that the radii of larger circles may en- compass almost all nodes of the network, and that the routing paths discovered by the al- gorithm do not have near-optimal hop counts (which may be important in quality of ser- vice applications). However, if the path quality is important, one can consider this algorithm only as the destination search step in the three-phase routing algorithm de- scribed above. A similar algorithm, using squares instead of circles and additional sophis- ticated techniques, is proposed in [31]. The location update techniques discussed so far include occasional flooding of location information to all or a large portion of nodes in the network. In the next two sections, methods that never use such flooding are discussed. 21.6 QUORUM-BASED STRATEGIES Quorum-based approaches for information dissemination are based on replicating infor- mation at multiple nodes acting as repositories. The choice of repositories and the query 460 LOCATION UPDATES FOR EFFICIENT ROUTING IN AD HOC NETWORKS Figure 21.2 Routing from S toward DЈ, DЈЈ, and D. [...]... breaking of some edges (indicated by arrows, with nodes A1, A2, A3, and A4) At position D3, it decides to send a location update in its current “column” by sending its position in the northern direction (it does so in the southern direction as well, but there is no neighbor in that direction in Figure 21.3) The main update path is indicated in bold lines, and bold, long dashed lines indicate some other nodes... routing for wireless networks, Proceedings MOBICOM, August, 2000, pp 243–254 27 G Karumanchi, S Muralidharan, and R Prakash, Information dissemination in partitionable mobile ad hoc networks, Proceedings IEEE Symposium on Reliable Distributed Systems, Lausanne, Oct., 1999 28 E Kranakis, H Singh, and J Urrutia, Compass routing on geometric networks, Proceedings 11th Canadian Conference on Computational... destination information that may be no more difficult to find than the other nodes in the same quorum A different quorum-based strategy, which deals with network dynamics, is proposed in [53] 462 LOCATION UPDATES FOR EFFICIENT ROUTING IN AD HOC NETWORKS In [53], nodes in an ad hoc network do not stay in the same “column,” and the distributed information may easily disperse due to node movement Moreover,... coordinates in the interval [0 m] Subgraphs can be used if obstacles are taken into account The connectivity depends on the selected transmission radius R Since transmission radius R for a given piece of equipment is normally 21.8 PERFORMANCE EVALUATION ISSUES 467 fixed (or should be selected from a few discrete values), most papers use a fixed value of R and change m to evaluate graphs of different... north–south direction, including neighbors of nodes in that path A initiates two routing messages, in the directions north and south, whereas other nodes follow only one of the directions Each follows variation of the MFR algorithm [55], with destination always to the north (or south) of the current node, as follows Current node B transmits update information to all its neighbors, and indicates, in... 21–24, 2000, Toronto, pp 173–180 50 I Stojmenovic, M Seddigh, and J Zunic, Internal node based broadcasting in wireless networks, Proceedings IEEE Hawaii International Conference on System Sciences, January 2001 51 W Su, S J Lee, M Gerla, Mobility prediction in wireless networks, Proceedings IEEE MILCOM, October, 2000 52 I Stojmenovic, Voronoi diagram and convex hull based geocasting and routing in... Optimal transmission ranges for randomly distributed packet radio terminals, IEEE Trans on Communications, 32, 3, 246–257, 1984 56 K Wu and J Harms, Location trace aided routing in mobile ad hoc networks, Proceedings IEEE ICCCN, Las Vegas, Oct., 2000 57 J Wu and H Li, On calculating connected dominating set for efficient routing in ad hoc wireless networks, Proceedings DIAL M, Seattle, Aug., 1999, pp 7–14... packet radio networks, IEEE Transactions on Communications, 34, 1, 38–44, 1986 18 Z J Haas and B Liang, Ad hoc mobility management with uniform quorum systems, ACM/IEEE Transactions on Networks, 7, 2, 228–240, 1999 19 Z J Haas and B Liang, Ad-hoc mobility management with randomized database groups, Proceedings of IEEE ICC, Vancouver, June, 1999 20 C Ho, K Obraczka, G Tsudik, and K Viswanath, Flooding for... networks They studied the problem of getting the location of some other node in the network and the surroundings of that node (e.g., firefighter) without the need to route any message to that node Their performance evaluation is limited in measuring the accuracy of the obtained information (i.e., the distance between the found and exact location of the other node) In [27], n nodes are divided into n1/2... system, Proceedings IEEE Wireless Communications and Networking Conference, New Orleans, Sept., 1999 4 S Basagni, I Chlamtac, V R Syrotiuk, B A Woodward, A distance routing effect algorithm for mobility (DREAM), Proceedings MOBICOM, 1998, pp 76–84 5 J Broch, D A Maltz, D B Johnson, Y C Hu, and J Jetcheva, A performance comparison of multihop wireless ad hoc network routing protocols, Proceedings MOBICOM, . framework with the corresponding notions in GEDIR and MFR methods. Stoj- menovic [52] discusses the V-GEDIR, CH-MFR, and R-DIR methods, in which m is for- warded. messages, as suggested by Lin and Liu [32]. They discussed this difference and even proposed an extreme difference in radii for short and long messages. They found